Research map / Vision-language and multimodal / Multimodal LLMs
HYDRA: A Hyper Agent for Dynamic Compositional Visual Reasoning
Fucai Ke, Zhixi Cai, Simindokht Jahangard, Weiqing Wang, Pari Delir Haghighi, Hamid Rezatofighi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add RL agent controller to ViperGPT-style pipeline (ablation) | ViperGPT · GQA | accuracy (%): 37.94 -> 43.71 | Improved |
| add RL agent on top of IR and sampling (ablation) | HYDRA · GQA | accuracy (%): 46.93 -> 47.88 | Improved |
| add incremental reasoning (IR) (ablation) | HYDRA · GQA | accuracy (%): 45.98 -> 47.88 | Improved |
| add multiple instruction sampling (ablation) | HYDRA · GQA | accuracy (%): 47.07 -> 47.88 | Improved |
| sample multiple instructions without RL agent (ablation) | ViperGPT · GQA | accuracy (%): 37.94 -> 39.84 | Improved |
| train RL agent on a different dataset (GQA instead of OK-VQA) (ablation) | HYDRA · OK-VQA | accuracy (%): 48.63 -> 48.17 | No clear effect |
| train RL agent on OK-VQA instead of A-OKVQA (ablation) | HYDRA · A-OKVQA | accuracy (%): 56.35 -> 55.94 | No clear effect |
| increase instruction samples to 10 (ablation) | HYDRA RL agent · OK-VQA | accuracy (%) | Hurt |
| use 5 instruction samples (ablation) | HYDRA · OK-VQA | accuracy (%): 48.63 | Improved |
| replace BLIP2 with LLaVA1.5 (7B) as VFM in HYDRA | HYDRA · GQA | accuracy (%): 47.9 -> 64.5 | Improved |
| wrap LLaVA1.5 (7B) in HYDRA | LLaVA1.5 (7B) · GQA | accuracy (%): 62.0 -> 64.5 | Improved |
| replace ViperGPT with HYDRA for visual grounding | ViperGPT · RefCOCO+ | IoU (%): 60.0 -> 61.1 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- BLIP-2 — BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation (ICML 2022) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MATA: A Trainable Hierarchical Automaton System for Multi-Agent Visual Reasoning (ICLR 2026 · similarity 0.90)
- ATLAS: Agentic or Latent Visual Reasoning? One Word is Enough for Both (NeurIPS 2026 · similarity 0.90)
- Insight-V: Exploring Long-Chain Visual Reasoning with Multimodal Large Language Models (CVPR 2025 · similarity 0.90)
- Unveiling the Compositional Ability Gap in Vision-Language Reasoning Model (NeurIPS 2025 · similarity 0.90)
- NePTune: A Neuro-Pythonic Framework for Tunable Compositional Reasoning on Vision-Language (ICLR 2026 · similarity 0.89)
- Multi-step Visual Reasoning with Visual Tokens Scaling and Verification (NeurIPS 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
